zhyang-dev/OpenFOAM-12
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

parallelProcessing: Various improvements

Browse Source 
boundaryProcAddressing has been removed. This has not been needed for a
long time. decomposePar has been optimised for mininum IO, rather than
minimum memory usage. decomposePar has also been corrected so that it
can decompose sequences of time-varying meshes.
This commit is contained in:
Will Bainbridge
2022-03-10 11:24:21 +00:00
parent dc052dd20c
commit 3995456979
26 changed files with 1582 additions and 1808 deletions
Download Patch File Download Diff File Expand all files Collapse all files

77
applications/utilities/parallelProcessing/decomposePar/fvFieldDecomposerDecomposeFields.C
View File

@ -2,7 +2,7 @@
========= |
\\ / F ield | OpenFOAM: The Open Source CFD Toolbox
\\ / O peration | Website: https://openfoam.org
\\ / A nd | Copyright (C) 2011-2021 OpenFOAM Foundation
\\ / A nd | Copyright (C) 2011-2022 OpenFOAM Foundation
\\/ M anipulation |
-------------------------------------------------------------------------------
License
@ -75,17 +75,17 @@ Foam::fvFieldDecomposer::decomposeField
) const
{
// 1. Create the complete field with dummy patch fields
PtrList<fvPatchField<Type>> patchFields(boundaryAddressing_.size());
PtrList<fvPatchField<Type>> patchFields(procMesh_.boundary().size());
forAll(boundaryAddressing_, patchi)
forAll(procMesh_.boundary(), procPatchi)
{
patchFields.set
(
patchi,
procPatchi,
fvPatchField<Type>::New
(
calculatedFvPatchField<Type>::typeName,
procMesh_.boundary()[patchi],
procMesh_.boundary()[procPatchi],
DimensionedField<Type, volMesh>::null()
)
);
@ -120,33 +120,38 @@ Foam::fvFieldDecomposer::decomposeField
typename GeometricField<Type, fvPatchField, volMesh>::
Boundary& bf = resF.boundaryFieldRef();
forAll(bf, patchi)
forAll(bf, procPatchi)
{
const fvPatch& procPatch = procMesh_.boundary()[patchi];
const fvPatch& procPatch = procMesh_.boundary()[procPatchi];
label fromPatchi = boundaryAddressing_[patchi];
if (fromPatchi < 0 && isA<processorCyclicFvPatch>(procPatch))
// Determine the index of the corresponding complete patch
label completePatchi = -1;
if (procPatchi < completeMesh_.boundary().size())
{
completePatchi = procPatchi;
}
else if (isA<processorCyclicFvPatch>(procPatch))
{
const label referPatchi =
refCast<const processorCyclicPolyPatch>
(procPatch.patch()).referPatchID();
if (field.boundaryField()[referPatchi].overridesConstraint())
{
fromPatchi = boundaryAddressing_[referPatchi];
completePatchi = referPatchi;
}
}
if (fromPatchi >= 0)
if (completePatchi != -1)
{
bf.set
(
patchi,
procPatchi,
fvPatchField<Type>::New
(
field.boundaryField()[fromPatchi],
field.boundaryField()[completePatchi],
procPatch,
resF(),
*patchFieldDecomposerPtrs_[patchi]
patchFieldDecomposers_[procPatchi]
)
);
}
@ -154,12 +159,12 @@ Foam::fvFieldDecomposer::decomposeField
{
bf.set
(
patchi,
procPatchi,
new processorCyclicFvPatchField<Type>
(
procPatch,
resF(),
(*processorVolPatchFieldDecomposerPtrs_[patchi])
processorVolPatchFieldDecomposers_[procPatchi]
(
field.primitiveField()
)
@ -170,12 +175,12 @@ Foam::fvFieldDecomposer::decomposeField
{
bf.set
(
patchi,
procPatchi,
new processorFvPatchField<Type>
(
procPatch,
resF(),
(*processorVolPatchFieldDecomposerPtrs_[patchi])
processorVolPatchFieldDecomposers_[procPatchi]
(
field.primitiveField()
)
@ -186,7 +191,7 @@ Foam::fvFieldDecomposer::decomposeField
{
bf.set
(
patchi,
procPatchi,
new emptyFvPatchField<Type>
(
procPatch,
@ -213,10 +218,6 @@ Foam::fvFieldDecomposer::decomposeField
const GeometricField<Type, fvsPatchField, surfaceMesh>& field
) const
{
// Apply flipping to surfaceScalarFields only
const bool doFlip = (pTraits<Type>::nComponents == 1);
// Problem with addressing when a processor patch picks up both internal
// faces and faces from cyclic boundaries. This is a bit of a hack, but
// I cannot find a better solution without making the internal storage
@ -243,17 +244,17 @@ Foam::fvFieldDecomposer::decomposeField
// 1. Create the complete field with dummy patch fields
PtrList<fvsPatchField<Type>> patchFields(boundaryAddressing_.size());
PtrList<fvsPatchField<Type>> patchFields(procMesh_.boundary().size());
forAll(boundaryAddressing_, patchi)
forAll(procMesh_.boundary(), procPatchi)
{
patchFields.set
(
patchi,
procPatchi,
fvsPatchField<Type>::New
(
calculatedFvsPatchField<Type>::typeName,
procMesh_.boundary()[patchi],
procMesh_.boundary()[procPatchi],
DimensionedField<Type, surfaceMesh>::null()
)
);
@ -282,7 +283,7 @@ Foam::fvFieldDecomposer::decomposeField
faceAddressing_,
procMesh_.nInternalFaces()
),
doFlip
isFlux(field)
),
patchFields
)
@ -296,21 +297,21 @@ Foam::fvFieldDecomposer::decomposeField
typename GeometricField<Type, fvsPatchField, surfaceMesh>::
Boundary& bf = resF.boundaryFieldRef();
forAll(boundaryAddressing_, patchi)
forAll(procMesh_.boundary(), procPatchi)
{
const fvPatch& procPatch = procMesh_.boundary()[patchi];
const fvPatch& procPatch = procMesh_.boundary()[procPatchi];
if (boundaryAddressing_[patchi] >= 0)
if (procPatchi < completeMesh_.boundary().size())
{
bf.set
(
patchi,
procPatchi,
fvsPatchField<Type>::New
(
field.boundaryField()[boundaryAddressing_[patchi]],
field.boundaryField()[procPatchi],
procPatch,
resF(),
*patchFieldDecomposerPtrs_[patchi]
patchFieldDecomposers_[procPatchi]
)
);
}
@ -319,7 +320,7 @@ Foam::fvFieldDecomposer::decomposeField
// Do our own mapping. Avoids a lot of mapping complexity.
bf.set
(
patchi,
procPatchi,
new processorCyclicFvsPatchField<Type>
(
procPatch,
@ -328,7 +329,7 @@ Foam::fvFieldDecomposer::decomposeField
(
allFaceField,
procPatch.patchSlice(faceAddressing_),
doFlip
isFlux(field)
)
)
);
@ -338,7 +339,7 @@ Foam::fvFieldDecomposer::decomposeField
// Do our own mapping. Avoids a lot of mapping complexity.
bf.set
(
patchi,
procPatchi,
new processorFvsPatchField<Type>
(
procPatch,
@ -347,7 +348,7 @@ Foam::fvFieldDecomposer::decomposeField
(
allFaceField,
procPatch.patchSlice(faceAddressing_),
doFlip
isFlux(field)
)
)
);